Re: [SQU] How to tell if Squid is Overloaded?

On Thu, 30 Nov 2000, Dana Gourley wrote:
 
> I found in the cache manager the "Store IO Interface Stats" page which seems
> to be what you might be talking about. It had two counters for
> "create.select_fail" and "create.success". When I watch these numbers
> durring a polygraph test at a high request rate there are alot of
> create.select_fails. Can anyone explain what the significance of these
> numbers are? "create.select_fail" doesn't sound good what ever it is.

The problem is in the function storeDirSelectSwapDirLeastLoad()
This function fails to select any swap directory if it thinks
all of them are too loaded.

As an alternative you can use round-robin selection by setting
store_dir_select_algorithm in squid.conf:

        store_dir_select_algorithm round-robin

That uses storeDirSelectSwapDirRoundRobin() instead, which always
selects a swap directory; it never fails.

> Haven't yet been able to see how these numbers might correlate to a lower
> hit ratio. There's an awful lot of create.select_fails right from the
> start.

storeDirSelectSwapDirLeastLoad is being called when you have an
object to swap out. When it fails (i.e. returns -1), the object
does not get swapped out. Thus, next time its requested polygraph
expects a hit, but it gets a miss.

Since you get a lot of "create.select_fails" it means you're pushing
Squid too hard. If you use round-robin with the same offered load,
your hit ratio may stay high, but I'll bet the response time goes
way up.

Duane W.

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html